Journal of Hydrology 526 (2015) 314

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Hydrological drought severity explained by climate and catchment

characteristics
A.F. Van Loon a,, G. Laaha b
a
b

Hydrology and Quantitative Water Management Group, Wageningen University, P.O. Box 47, 6700 AA Wageningen, The Netherlands
Institute of Applied Statistics and Computing, University of Natural Resources and Life Sciences, BOKU Vienna, Gregor-Mendel Strasse 33, 1180 Vienna, Austria

a r t i c l e

i n f o

Article history:
Available online 22 November 2014
Keywords:
Hydrological drought severity
Drought duration
Drought decit
Climate
Catchment characteristics
Storage

s u m m a r y
Impacts of a drought are generally dependent on the severity of the hydrological drought event, which can
be expressed by streamow drought duration or decit volume. For prediction and the selection of drought
sensitive regions, it is crucial to know how streamow drought severity relates to climate and catchment
characteristics. In this study we investigated controls on drought severity based on a comprehensive
Austrian dataset consisting of 44 catchments with long time series of hydrometeorological data (on average
around 50 year) and information on a large number of physiographic catchment characteristics. Drought
analysis was performed with the variable threshold level method and various statistical tools were applied,
i.e. bivariate correlation analysis, heatmaps, linear models based on multiple regression, varying slope
models, and automatic stepwise regression. Results indicate that streamow drought duration is primarily
controlled by storage, quantied by the Base Flow Index or by a combination of catchment characteristics
related to catchment storage and release, e.g. geology and land use. Additionally, the duration of dry spells
in precipitation is important for streamow drought duration. Hydrological drought decit, however, is
governed by average catchment wetness (represented by mean annual precipitation) and elevation
(reecting seasonal storage in the snow pack and glaciers). Our conclusion is that both drought duration
and decit are governed by a combination of climate and catchment control, but not in a similar way.
Besides meteorological forcing, storage is important; storage in soils, aquifers, lakes, etc. inuences drought
duration and seasonal storage in snow and glaciers inuences drought decit. Consequently, the spatial
variation of hydrological drought severity is highly dependent on terrestrial hydrological processes.
2014 The Authors. Published by Elsevier B.V. All rights reserved.

1. Introduction
Drought is considered one of the most damaging natural disasters in terms of economic costs (e.g. navigation and hydropower
production, Wilhite, 2000; Carroll et al., 2009; Van Vliet et al.,
2012), societal problems (e.g. increased mortality and conict,
Garcia-Herrera et al., 2010; Hsiang et al.) and ecological impacts
(e.g. forest dieback and impacts on aquatic ecosystems, Lake,
2011; Lewis et al., 2011; Choat et al., 2012). Drought is commonly
dened as a below-normal water availability (Wilhite and Glantz,
1985; Wilhite, 2000; Tallaksen and Van Lanen, 2004; Shefeld
and Wood, 2011; Mishra and Singh, 2010), but there is no real consensus about the application of this denition (Hayes et al., 2010).
In this study we assume that society and the ecosystem are
adapted to the seasonal cycle and we regard drought as a deviation
from this seasonal cycle, which means that droughts also occur in
Corresponding author at: School of Geography, Earth and Environmental
Sciences, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK.
E-mail addresses: anne.vanloon@wur.nl, a.f.vanloon@bham.ac.uk (A.F. Van Loon).
http://dx.doi.org/10.1016/j.jhydrol.2014.10.059
0022-1694/ 2014 The Authors. Published by Elsevier B.V. All rights reserved.

the high ow season. Drought is subdivided into different types of

drought related to the variables of the hydrological cycle, precipitation (meteorological drought), soil moisture (soil moisture
drought), and groundwater and streamow (hydrological drought)
(Tallaksen and Van Lanen, 2004). Almost all drought impacts are
related to soil moisture drought or hydrological drought, since
both the ecosystem and society depend upon water from the
catchment stores (soil, aquifers, lakes, rivers) rather than from precipitation directly. Hydrological drought is determined by the
propagation of meteorological drought through the terrestrial
hydrological cycle and is therefore inuenced by the properties
of the hydrological cycle (Peters et al., 2006; Van Lanen, 2006;
Vidal et al., 2010). For example, drought propagation is different
in an semi-arid climate and a climate with snow accumulation
in winter, and it differs between mountainous catchments,
catchments with many lakes and wetlands, and catchments with
mild slopes and large, porous aquifers (Van Loon, 2013).
Besides drought frequency (how often a drought occurs),
drought severity (the strength of a drought) is an important characteristic of drought events since it is directly related to the

A.F. Van Loon, G. Laaha / Journal of Hydrology 526 (2015) 314

impacts of drought (Hayes et al., 2010). Drought severity can be

quantied in various ways. In standardised indices (e.g. Standardised Precipitation Index, SPI, McKee et al., 1993), and Standardised
Groundwater level Index, SGI (Bloomeld and Marchant, 2013),
which are increasingly used in scientic drought studies (e.g.
Vicente-Serrano et al., 2009; Mishra et al., 2009; Joetzjer et al., 2013),
drought severity is expressed by the number of standard deviations from the mean. For most impacts, however, more physical
measures of severity are needed (Wong et al., 2013). For many
aquatic ecosystems for example the duration of a drought in
streamow is crucial (Bond et al., 2008), whereas for hydropower
production the missing volume of water compared to normal conditions (decit volume) is more relevant (Jonsdottir et al., 2005;
Rossi et al., 2012; Tsakiris et al., 2013).
Hydrological drought duration and decit are related since the
decit accumulates over the duration of the drought event (e.g.
Dracup et al., 1980; Woo and Tariiule, 1994; Shiau and Shen,
2001; Kim et al., 2003; Hisdal et al., 2004; Mishra et al., 2009;
Wong et al., 2013). Van Lanen et al. (2013) and Van Loon et al.
(2014) have shown that this relation is not linear. It is dependent
on propagation of the drought (Van Loon et al., 2014) and relates
strongly to climate and catchment characteristics (Van Lanen
et al., 2013). Van Lanen et al. (2013) assessed the effect of climate
(Kppen classes), soil and groundwater system on the bivariate
probability distribution of drought duration and decit. They found
that the responsiveness of the groundwater system is as important
for hydrological drought development as climate.
What is still unclear is how hydrological drought duration and
decit relate to climate and catchment characteristics and which
factor is dominant. Tallaksen and Hisdal (1997) speculated that
The distribution of drought duration is primarily thought to be
governed by climate. However decit volume is expected to be
more related to catchment characteristics (Tallaksen and Hisdal,

1997). More recent studies however have shown convincingly that

in a given climate hydrological drought duration is strongly related
to the responsiveness of the groundwater system, both in a theoretical analysis and in a real world example (Peters et al., 2003;
Peters et al., 2005). On the other hand, there are indications of an
effect of climate on drought decit, because in many studies the
decit volume of hydrological drought is standardised by dividing
by mean discharge to be able to compare catchments with different wetness (Clausen and Pearson, 1995; Kjeldsen et al., 2000;
Van Lanen et al., 2013). A quantitative analysis of the effects of climate and catchment control on drought duration and decit has, to
our knowledge, never been done.
We intend to ll that gap and investigate the relative effects of
climate and catchment on hydrological drought duration and deficit volume. For this study we used an extensive Austrian dataset,
that contains observations of precipitation, temperature and discharge for a high number of catchments and includes thematic
information for each catchment, e.g. climate, elevation, geology,
land use (Laaha and Blschl, 2006; Gal et al., 2012; Haslinger
et al., 2014). By combining different types of analysis we hope to
prove whether climate or catchment properties are more
important in determining both drought duration and decit. In
Section 2, we will rst describe the study area and data availability.
Sections 3.1 and 4.1 deal with the drought analysis methods and its
results and Sections 3.2 and 4.2 with the statistical analysis
methods and its results. Finally, discussion and conclusions are
given in Sections 5 and 6.
2. Study areas
The study has been conducted on a comprehensive Austrian
dataset consisting of 44 catchments which are free from major
disturbances. The study area is quite divers and the catchments

Fig. 1. Some characteristics of the study area: (a) clusters, (b) topography, (c) geology (all from Gal et al. (2012), reprinted with permission from the publisher Wiley), and (d)
mean annual precipitation (from ZAMG).

A.F. Van Loon, G. Laaha / Journal of Hydrology 526 (2015) 314

reect a broad range of hydrological regimes. The catchments are

pooled into 13 clusters which are homogeneous in terms of climate
conditions and hydrological processes (Fig. 1a). The pooling is
based on the work of Gal et al. (2012) on ood time scales which
have been shown to be a rich ngerprint of the hydrological
processes in a catchment. In fact, the spatial pattern of the
hotspots (as they were termed in Gal et al. (2012)) is very similar
to that of the low-ow regions of Laaha and Blschl (2006),
discerned from seasonality analysis. The hotspots are on average
smaller and, hence, likely more homogeneous. They obviously represent major climate and geological units of Austria (Fig. 1c and d)
which are indeed also relevant for hydrological droughts
(Haslinger et al., 2014).
Some clusters belong to regions where minimum ows typically occur in summer in consequence of low seasonal precipitation and high seasonal evaporation. These are Innviertel (Innv),
Mhlviertel (Muhlv), Waldviertel (Waldv), Weinviertel (Weinv),
Flysch, and Leitha. The catchments in the Alps, i.e. ztal (Ozt), High
Alps (Hoalp), and to some degree Bregenzerwald (Brewa) and
Dachstein (Dachst), are situated at altitudes where snow storage
processes have a major inuence on the ow regime (Fig. 1b). They
give rise to highly seasonal regimes with minimum ows in winter
and high discharges in summer. Catchments in the south also
belong to summer low ow type, but exhibit a particular climate
as they are situated leeward of the Alps and are screened from
moist Atlantic air masses. This yields lower amounts of precipitation then in the north, and often longer dry spells. As can be seen
from Fig. 1d, clusters in the south (Gail, Gurk) and southeast
(Buwe), but also in southern High Alps (Ozt) are affected by this
particular climate. North of the Alps a precipitation gradient can
be observed, reecting increasing aridity towards Eastern Europe.
For each catchment, discharge time series were available from
the Hydrographical Service of Austria (HZB) in daily resolution for
different periods between 1951 and 2010. The length of the discharge series was on average 47 year with minimum 27 year and
maximum 60 year. A linear interpolation was applied to ll a small
number of gaps of a few days. We used the total length of the discharge time series to calculate hydrological drought characteristics
duration and decit (Section 3.1), which are the target variables of
this study. For each cluster, one close-by meteorological station
was selected which seems representative for its climatic conditions.
In case multiple stations were present within or in the vicinity of the
cluster, we chose the station that was most representative in terms
of topographical factors such as elevation and exposition, since both
precipitation and temperature are dependent on these factors. Considering exposition is notably important for Alpine catchments as
there are major climatic differences between northern and southern
slopes of the Alpine divide. For the selected stations daily temperature and precipitation data were available from the Central Institute
for Meteorology and Geodynamics (ZAMG) for a period of on average 60 year (ranging from 46 to 62 year per station). Daily precipitation data of these stations were then used to calculate the
meteorological drought characteristics per cluster.
For analysing the effect of catchment characteristics on hydrological drought we used the catchment dataset of Laaha and
Blschl (2006), consisting of 31 physiographic catchment characteristics (Table 1). They relate to catchment area (SUM.AREA),
topographic elevation (H), topographic slope (SL), precipitation
(N), geological classes (GEOL), land use classes (BONU), and stream
network density (SDENS). In addition, the Base Flow Index (BFI)
and Recession coefcient (Rec) were calculated by aid of the
R-package lfstat (Kofer and Laaha, 2013), based on the WMO
Manual on Low-ow Estimation and Prediction (WMO, 2008). BFI
and Rec have been shown to reect storage and release properties
of the catchments (e.g. Salinas et al., 2013) and are therefore used
as catchment characteristics in this study.

Table 1
Catchment characteristics (based on Laaha and Blschl (2006)).
Name

Variable

Unit

SUM.AREA

Subcatchment area

101 km

H.MIN

Altitude of stream gauge

102 m

H.MAX

Maximum altitude

102 m

H.DIFF

Range of altitude

102 m

H.MEAN

Mean altitude

M.NEIG
SL.FL
SL.MG
SL.ST
N.GES

Mean slope
Slight slope
Moderate slope
Steep slope
Average annual precipitation

102 m
%
%
%
%

N.SOM

Average summer precipitation

102 mm

N.WIN

Average winter precipitation

GEOL.BM
GEOL.QUA
GEOL.TER
GEOL.FLY
GEOL.KAL
GEOL.KRI
GEOL.SHAL
GEOL.DEEP
GEOL.QUELL
BONU.URB
BONU.ACK
BONU.DAU
BONU.GRU
BONU.WAL
BONU.FEU
BONU.LOS
BONU.WAS
BONU.EIS
SDENS

Bohemian Massif
Quaternary sediments
Tertiary sediments
Flysch
Limestone
Crystalline rock
Shallow groundwater table
Deep groundwater table
Source region
Urban
Agriculture
Permanent crop
Grassland
Forest
Wetland
Wasteland (rocks)
Water surfaces
Glacier
Stream network density

102 mm
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%

BFI
Rec

Baseow index
Recession coefcient

102 mm

102 m km

day

3. Methodology
3.1. Drought analysis
Droughts were identied from the time series of precipitation
and discharge with the widely-used threshold level approach
(Zelenhasic and Salvai, 1987; Hisdal et al., 2004; Fleig et al.,
2006; Van Loon, 2013), dening a drought when a variable falls
below a pre-dened threshold. To reect seasonality we used a
variable threshold based on the 80th percentile of the ow duration curves of a 30 days moving window (Beyene et al., 2014).
This means that every day in the year has a different threshold
level based on the 80th percentile of the ow duration curve of
the discharge/precipitation measured on that day, the 15 days
before that day, and the 15 days after that day, for all years in
the time series. This method has proven to be most robust in
catchments with pronounced seasonality, for example catchments
dominated by snow accumulation and melt (Beyene et al., 2014).
We applied a pooling procedure to both time series of precipitation and discharge to pool dependent drought events, namely a
30-day moving average (Tallaksen and Hisdal, 1997; Hisdal
et al., 2004; Fleig et al., 2006). The duration of a drought event
was determined by calculating the total number of consecutive
days that the variable was below the threshold and the decit
volume is the sum of the deviations from the threshold times
the number of days (so the area between the two curves when
the variable is below the threshold). Minor droughts with a duration of less than 3 days were removed. Subsequently, the statistics
mean, maximum and standard deviation were calculated for both
drought duration and decit.

A.F. Van Loon, G. Laaha / Journal of Hydrology 526 (2015) 314

3.2. Statistical analysis

We investigated the effects of climate and catchment characteristics on discharge drought duration and decit by statistical
analysis. In the rst part of the investigation we analysed the
strength of the relationships between discharge drought and
the individual climate and catchment characteristics using bivariate
correlation analysis. We calculated the correlation matrix of pairwise combinations of all variables based on Pearson correlation
coefcients. The correlation matrix allowed us to nd important
relationships between discharge drought characteristics and the
various climate and catchment characteristics. As relationships
could be non-linear, we also computed Spearman correlation
coefcients and studied correlation plots to verify the results.
Correlation analysis is limited by the fact that catchment characteristics are likely not independent, due to co-evolution of
landscape, climate and geology. To explore intercorrelations of
catchment characteristics, we analysed the correlation structure
using heatmaps as implemented in the R software package stats.
Heatmaps employ an appropriate colour-coding to visualise sign
and strength of relationships from the correlation matrix.
Moreover, the algorithm performs a clustering of the pairwise
correlations in order to nd groups of variables which are dependent on each other. After rows and columns are arranged
according to similarity, previously undetectable patterns can
become obvious. The analysis is therefore useful for nding
groups of climate and catchment characteristics which have a
joint effect on drought characteristics. Based on hydrological
reasoning, we aim to identify key variables of each group which
appear as the most important physical factors in drought
generation.
In the second step we extended the scope of analysis to
interactions of key variables in drought generation. For this we
applied linear models with multiple regression and interaction
terms, which make the linear model more exible to represent
possible non-linear relationships. We also applied varying slope
models to perform tting individually for different climatological
regions. We tested important combinations of predictor variables
as suggested by correlation analysis and compared them with
automatic stepwise (combined forward and backward) regression
analysis based on the Akaike Information Criterion (Akaike,
1974). If the model appears well representative (from residual
statistics and in terms of model assumptions) it can be used to
analyse the effects of a predictor on the drought characteristic.
By effect we denote the change in a response variable produced
by a change in one or more explanatory or factor variables,
adjusted for the other variables in the model. In case of more
than one predictor, the effect of one explanatory variable (regression line) is adjusted for the other variables in the model. This is
similar to partial correlation analysis where the correlation
between two variables is adjusted for the effect of a confounding
variable. Linear model effects, however, are more general than
partial correlations, since the method allows a simultaneous
consideration of a number of predictors including possible
interactions. The computation and visualisation of the effects of
the regression terms is conducted using the R-package effects
(Fox, 2003).

meteorological droughts. The largest difference is found between

Gail and Inn (Gail: 4.5 droughts per year with an average duration
of 16.2 days and Inn: 5.2 droughts per year with an average duration of 13.4 days). This difference is related to the variability in precipitation described in Section 2. In Gail there are less, but longer
dry spells, whereas in Inn the higher number of short dry spells
points to a larger variability in precipitation. The decit volume
of droughts in precipitation is more variable with 3.9 mm for
Weinv and 13.3 mm for Brewa. This difference is due to differences
in mean annual precipitation, which amounts to 500 mm per year
for Weinv and 1900 mm per year for Brewa. A higher mean annual
precipitation results in a higher threshold, which produces higher
decit volumes.
In all catchments we see a clear propagation of the drought signal from meteorological to hydrological drought (Tables 2 and 3):
there are fewer but longer droughts in discharge than in precipitation (on average 2.2 droughts per year with an average duration of
35 days in discharge vs. 4.8 droughts per year with an average
duration of 15 days in precipitation). There is much more variability between the clusters in hydrological drought characteristics
than in meteorological drought characteristics. The number of
discharge droughts varies between 1.4 and 3.2 and average duration ranges from 23 to 54 days. Gurk has fewest and longest
droughts and Brewa has most and shortest droughts. The average
decit volume of droughts in discharge is comparable with that
of droughts in precipitation, but the ranges are even larger
(0.518 mm per cluster and 0.322 mm per catchment). Some
clusters are homogeneous in terms of discharge drought
Table 2
General drought characteristics using a 80 % variable threshold (based on a moving
window of 30 days), the moving average method for pooling, and a minimum
duration of 3 days for the hydrometeorological variables of all clusters.
Station

No. of
droughts
[per
year]

Mean
duration
[day]

Max
duration
[day]

Mean
decit
[mm]

Max
decit
[mm]

Brewa

Precipitation
200154
200204
200287
Average Q

4.8
3.1
3.5
3.1
3.2

15.1
24
21
23
23

85
144
130
128
134

13.3
17
16
22
18

133
284
226
398
303

Buwe

Precipitation
208827
208835
210245
208835
208827
Average Q

4.8
1.4
1.5
1.4
1.6
1.9
1.6

15.0
53
48
53
45
40
48

95
409
321
437
311
426
381

6.8
5.2
4.9
3.0
3.2
3.8
4.0

57
51
47
48
42
92
56

Dachst

Precipitation
205799
205831
205856
210583
Average Q

4.7
2.8
2.6
2.7
2.8
2.7

14.8
26
28
27
26
27

73
143
143
122
142
138

12.5
14
13
8.2
17
13

99
186
138
89
223
159

Gail

Precipitation
212613
212647
212670
212753
Average Q

4.5
1.8
1.4
1.6
1.7
1.6

16.2
42
53
46
42
46

150
301
293
213
216
256

10.7
14
18
15
13
15

113
166
123
117
115
130

Gurk

Precipitation
212860
212951
Average Q

4.9
1.7
1.1
1.4

14.7
43
66
54

76
254
379
317

9.1
8.9
6.4
7.6

78
70
42
56

Hoalp

Precipitation
212068
212076
Average Q

4.9
2.9
2.9
2.9

14.1
25
25
25

92
147
141
144

7.4
19
16
17

81
174
175
175

4. Results
4.1. Drought analysis
Drought analysis on the precipitation data (Tables 2 and 3)
shows that there is little difference between clusters in the number
(around 4.8 per year) and average duration (around 15 days) of

A.F. Van Loon, G. Laaha / Journal of Hydrology 526 (2015) 314

Table 3
Continuation - General drought characteristics using a 80% variable threshold (based
on a moving window of 30 days), the moving average method for pooling, and a
minimum duration of 3 days for the hydrometeorological variables of all clusters.
Max
decit
[mm]

Station

No. of
droughts
[per year]

Mean
Max
Mean
duration duration decit
[mm]
[day]
[day]

Innviertel

Precipitation
204768
204784
204834
204859
204958
205047
Average Q

5.2
2.8
2.5
2.7
2.7
2.4
2.4
2.6

13.4
26
29
27
26
31
30
28

79
202
285
180
181
158
149
193

5.6
3.8
4.9
3.5
3.7
3.2
3.5
3.8

57
47
66
27
31
32
34
40

Leitha

Precipitation
208413
209007
210013
Average Q

5.0
1.6
2.0
1.9
1.8

14.4
46
38
42
42

90
286
256
245
262

4.6
4.7
1.9
1.6
2.8

40.9
32
20
20
24

Muhlviertel Precipitation
204875
204891
204917
204925
Average Q

4.6
2.2
2.1
2.1
1.8
2.0

15.2
34
35
34
40
36

91
234
242
186
205
217

6.7
6.7
8.3
7.1
5.7
7.0

60
78
66
70
40
64

Ozt

Precipitation
201350
201376
201392
201418
201434
Average Q

5.0
2.3
2.4
2.5
2.4
2.7
2.4

13.9
28
26
25
24
24
25

161
222
161
205
193
192
195

6.0
14.0
14.3
9.3
7.0
6.5
10.2

94
203
205
120
99
92
144

Waldviertel Precipitation
205997
207944
Average Q

4.9
2.1
2.2
2.1

14.8
31.2
27.5
29

80
246
256
251

4.6
4.5
2.6
4

64
44.2
35.1
40

Weinviertel Precipitation
208041
208058
208447
208637
208678
209189
Average Q

4.8
1.8
1.5
2.5
2.0
2.0
2.2
2.0

15.1
42
48
29
36
39
33
38

76
352
391
133
348
295
198
286

3.9
0.59
0.53
0.37
0.36
0.71
0.32
0.5

28
6.7
7.4
2.5
5.0
11.2
2.6
6

characteristics, others are quite heterogeneous. Most variability is

visible in the Weinv cluster. The longest drought occurs in Buwe
(437 days, more than 14 months) and the maximum decit occurs
in Brewa (398 mm). The different values for average drought duration and decit in Tables 2 and 3 are a rst indication that drought
duration and decit might have different governing factors.
This is even more apparent from the spatial distribution of
mean duration and decit of both precipitation and discharge
droughts (Fig. 2). We can see that:
 the average duration of precipitation droughts (P_meandur) is
high in the East and South;
 the average decit volume of precipitation droughts (P_meandef) is highest in the northern Alps;
 the average duration of discharge droughts (Q_meandur) is high
in the East and South, but more scattered than P_meandur;
 the average decit volume of discharge droughts (Q_meandef)
is high in the southern Alps.
When comparing Fig. 2 with Fig. 1, some of the patterns can be
explained. The general patterns of both P_meandur and Q_meandur reect the inuence of climate: in the Alps, precipitation
events are more frequent than in the lowlands, so droughts in
the Alpine region have the lowest duration. In the region north

of the Alps, the patterns reect increasing drought duration from

west to east which is obviously related to decline in annual precipitation and increasing aridity (Fig. 1d). South and southeast of the
Alps longer dry spells are observed, because the catchments there
are screened by the Alps (Section 2). Overall, Q_meandur has
longer time scales than P_meandur, reecting drought propagation, i.e. pooling of shorter meteorological drought events into
longer streamow droughts as a result of catchment storage processes. Differences between catchments in the same cluster are
likely the effect of geology (storage, Fig. 1c), which is most
pronounced in the eastern (Weinv) and southern clusters (Gail
and Gurk). This would indicate a combined effect of climate and
catchment control on hydrological drought duration.
It is interesting to see that the patterns of drought decit in
precipitation and discharge (P_meandef and Q_meandef) are very
different from the patterns of drought duration (Fig. 2). Highest
decit volumes are found in the Alpine region. High decit
volumes in the northern Alps are expected because of the high
catchment wetness (Fig. 1d) resulting in higher threshold values,
but the high decit volumes in discharge (Q_meandef) seem to
reect more the spatial distribution of altitude (Fig. 1b) than that
of precipitation (Fig. 1d). So, the question arises whether there is
an additional effect of catchment properties related to elevation
on decit volume of hydrological droughts.
For a selection of clusters, we studied some catchment characteristics that were assumed to inuence drought duration and deficit (i.e. area, elevation, BFI and recession constant) in more detail
(Table 4). Some conclusions can be drawn:
 All Brewa catchments have many droughts with short duration and high decit (Table 2), they also have a small area,
high elevation, low BFIs and low recession constants. The
geology of the Brewa catchments is characterised as Flysch.
According to Gal et al. (2012) Flysch tends to produce very
ashy response as the ow paths are at the surface or very
near the surface with little inltration. This has important
implications for oods (short ood time scales, Gal et al.
(2012)), but also for drought.
 All Muhlv catchments have few droughts with long duration
and average decit (Table 3), they also have a large area, low
elevation, high BFIs and high recession constants.
 The Innv and Weinv catchments show high variation in area,
BFI and recession constant, whereas elevation is comparable.
This offers the possibility to separate the effect of storage
from climate and elevation effects. For Innv, catchments with
a high BFI, also have longer drought events and lower decit
volume (Table 3). For Weinv the relation is less clear.
4.2. Statistical analysis
Statistical analysis allows for a quantitative investigation of
relations between hydrological drought duration and decit and
possible governing factors (climate and catchment characteristics).
The heatmap in Fig. 3 shows the correlation structure between
average duration and decit of droughts in discharge (Q_meandur
and Q_meandef), average duration and decit of droughts in precipitation (P_meandur and P_meandef) and a selection of climate
and catchment characteristics from Table 1. We did not include
all variables because some are interchangeable in the sense that
they correlate the same way with other variables, for example
the variables related to altitude (H.MIN, H.MAX, H.MEAN) and precipitation (N.GES, N.SOM, N.WIN).
In Fig. 3 red squares indicate high positive correlation, blue
squares indicate high negative correlation. These correlations are
based on Pearson correlation. The heatmap based on Spearman
correlation coefcients to test for non-linearity showed a similar

A.F. Van Loon, G. Laaha / Journal of Hydrology 526 (2015) 314

(b)

(a)

17

49.0
48.5

49.0

13

48.5

16

11

48.0
15

47.5

lat

lat

48.0
9
47.5
7

47.0

47.0

14

5
46.5

46.5
13
10

12

14

16

10

12

lon

14

16

lon

(d)

(c)
49.0

70

49.0

48.5

60

48.5

48.0

50

48.0

47.5

40

47.0

15

lat

lat

20

47.5

10

47.0
5

30
46.5

46.5
0

20
10

12

14

16

10

lon

12

14

16

lon

Fig. 2. Drought duration and decit in precipitation and discharge. (a) P_meandur [days], (b) P_meandef [mm], (c) Q_meandur [days], (d) Q_meandef [mm]. Note the different
colour scales.

Table 4
Selected catchment characteristics per catchment for a subset of clusters (variable
names see Table 1).
SUM.AREA
2

[101 km ]
Brewa

[102 m]

H.MEAN

BFI
[]

Rec
[]

200154
200204
200287

33
54
31

1472
1159
1510

0.54
0.30
0.43

10
6.7
11

Innviertel

204768
204784
204834
204859
204958
205047

70
60
81
303
66
29

456
460
486
443
460
420

0.45
0.42
0.67
0.52
0.74
0.49

20
13
16
11
17
16

Muhlviertel

204875
204891
204917
204925

135
123
255
139

693
790
763
759

0.69
0.67
0.68
0.68

19
19
18
19

Weinviertel

208041
208058
208447
208637
208678
209189

213
380
130
370
69
515

337
285
273
321
232
260

0.72
0.77
0.69
0.60
0.79
0.77

16
19
18
10

pattern. This indicates that we can assume that in our case linear
models are also suited to represent monotonic relationships even
if they are not perfectly linear. The ordering of variables is the
result of hierarchical cluster analysis using Euclidean distance. This
yields that variables with similar correlation patterns are grouped
together. An example is the high correlation and close clustering
between SL.FL, BONU.ACK, GEOL.TER and BONU.URB, which indicates that urban area and agriculture are present in regions with

a slight slope and tertiary sediments. This is not surprising and

does not give any information explaining drought characteristics.
In the following paragraphs we only focus on factors that relate
to the duration and decit volume of discharge droughts.
4.2.1. Drought duration
The average duration of droughts in discharge (Q_meandur)
shows highest correlation (Fig. 3) with the baseow index (BFI).
From the scatterplot (Fig. 4a) this relation is obvious. BFI itself is
not a catchment characteristic but it integrates the effect of storage
and response times of a catchment (Section 2). The same is true for
the recession constant (Rec), but Rec has a lower correlation with
Q_meandur.
It is also interesting that Q_meandur has a much lower correlation with catchment characteristics indicative of catchment storage, such as area percentages of geological classes (GEOL.QUA,
GEOL.KRI), aquifers (GEOL.SHAL, GEOL.DEEP), lakes and wetlands
(BONU.WAS, BONU.FEU), and catchment area (SUM.AREA), than
with BFI (Fig. 3). Although all these correlations are plausible
regarding their sign (characteristics reecting high storage are positively correlated), not one of them seems dominant. One of the
reasons for that is that not all catchments have data for all variables. In Fig. 4a, for example, all catchments with some degree of
GEOL.DEEP and GEOL.SHAL are indicated. These correspond to high
BFI, but the low number of catchments with GEOL.DEEP or
GEOL.SHAL makes nding a quantitative relation difcult.
After BFI, the highest correlation of Q_meandur is with P_meandur, and there is also a strong negative correlation with mean
annual precipitation (N.GES) (Fig. 3). This points at an effect of
climate on the duration of droughts in discharge. In the drier catchments with longer dry spells the duration of discharge droughts is
longer (Fig. 4b), but the relation is less clear than with BFI.

A.F. Van Loon, G. Laaha / Journal of Hydrology 526 (2015) 314

SL.FL
BONU.ACK
GEOL.TER
BONU.URB
SDENS
GEOL.BM
GEOL.SHAL
Q_meandur
BFI
SUM.AREA
BONU.FEU
GEOL.QUA
GEOL.DEEP
BONU.WAS
H.MEAN
M.NEIG
SL.ST
BONU.LOS
BONU.EIS
RecMRC
GEOL.KRI
P_meandur
GEOL.KAL
BONU.WAL
Q_meandef
P_meandef
N.GES
SL.MG
BONU.GRU
GEOL.FLY
GEOL.QUELL

0.5

0.5

GEOL.QUELL
GEOL.FLY
BONU.GRU
SL.MG
N.GES
P_meandef
Q_meandef
BONU.WAL
GEOL.KAL
P_meandur
GEOL.KRI
RecMRC
BONU.EIS
BONU.LOS
SL.ST
M.NEIG
H.MEAN
BONU.WAS
GEOL.DEEP
GEOL.QUA
BONU.FEU
SUM.AREA
BFI
Q_meandur
GEOL.SHAL
GEOL.BM
SDENS
BONU.URB
GEOL.TER
BONU.ACK
SL.FL

Fig. 3. Heatmap of correlations between drought severity indices, climate and catchment characteristics. Euclidean distances used for clustering.

(a)

(b)

Brewa
Buwe
Dachst
Flysch
Gail
Gurk
Hoalp
Innv
Muhlv
Ozt
Waldv
Weivn

0.4

0.5

0.6

0.7

0.8

BFI

50

13.5

14.0

20

0.3

40

Q_meandur

50
40

60

Brewa
Buwe
Dachst
Flysch
Gail
Gurk
Hoalp
Innv
Muhlv
Ozt
Waldv
Weivn

30

20

30

Q_meandur

60

14.5

15.0

15.5

16.0

P_meandur

Fig. 4. Relation between the average duration of droughts in discharge (Q_meandur) and (a) the baseow index (BFI), colour coded per cluster. Catchments with deep or
shallow groundwater (GEOL.SHAL and GEOL.DEEP) are indicated with circles and (b) the average duration of droughts in precipitation (P_meandur).

To investigate the effects of climate and catchment characteristics on drought duration in more detail, we studied a number of
linear regression models. In Table 5 the three best models are
described. The model with only BFI is highly signicant (model
1) and the model with only P_meandur is also signicant (model
2). P_meandur also adds information to the model with BFI (model
3) and so much that this combined model is the best model to
explain Q_meandur. Adding more complex interactions or adding
other variables did not improve the model.

With the same analysis we can nd out if any combination of

catchment characteristics can replace BFI to explain Q_meandur.
In Table 5 we see that only using catchment characteristics does
not yield a signicant model (model 4). Adding P_meandur to
model 4 does yield a signicant model (model 5). There are differences in signicance between the different catchment characteristics, but again no factor seems dominant. All variables are needed
to yield a model that is comparable to the model with BFI and
P_meandur (model 3), excluding one of the variables from model

10

A.F. Van Loon, G. Laaha / Journal of Hydrology 526 (2015) 314

each of these characteristics alone is rather low as compared

to BFI, which is highly informative for streamow drought
duration.

Table 5
Linear models tested to explain the average duration of discharge droughts
(Q_meandur), with multiple regression and interaction terms, including their
statistical signicance.
Model 1

Model 2

Model 3

BFI
P_meandur

1.33e 07 ()

0.00103 ()

5.27e 08 ()
0.0277 ()

Comparing models

1.001e 06 ()

Model 4

Model 5

0.1618
0.0172
0.0857
0.0926
0.0152
0.0813
0.6750
0.3715
0.6193

7.14e 05 ()
0.20327
0.00629 ()
0.04168 ()
0.00613 ()
0.66517
0.02959 ()
0.12049
0.03521 ()
0.79736

Model 3

BFI
P_meandur
GEOL.SHAL
GEOL.DEEP
GEOL.QUA
GEOL.KRI
BONU.WAL
BONU.WAS
BONU.FEU
SDENS
SUM.AREA

5.27e 08 (
0.0277 ()

Comparing models

7.281e 06 ()

( )
(.)
(.)
( )
(.)

4.2.2. Drought decit

The average decit of droughts in discharge (Q_meandef) shows
high correlation (Fig. 3) with variables related to climate (N.GES
and P_meandef) and catchment characteristics (H.MEAN, M.NEIG,
SL.ST, BONU.EIS, BONU.LOS, GEOL.KAL). These variables are not
independent: precipitation generally increases with altitude,
slopes are higher in the Alps, glaciers are present only above
2500 m, and calcareous rocks are found in the northern Alps
(Fig. 1). All these characteristics are nally related to catchment
altitude. To disentangle the effects of climate and altitude we studied the relation of Q_meandef with mean annual precipitation
(N.GES) and mean elevation (H.MEAN). Both show a positive relation (Fig. 5a and b). The relation with N.GES was expected; high
average precipitation results in high average discharge and therefore a high threshold. Fluctuations around the threshold then
result in high decit volumes, as can be seen in Fig. 6a, in which
the hydrographs of two catchments in the same cluster are compared. The catchments show similar drought periods but decit
volumes are different, i.e. the upper (208413) and lower
(209007) catchment have an average discharge of 1.4 and
0.5 mm/d and an average decit of 4.7 and 1.9 mm, respectively
(Table 3). Hence, catchments with higher mean discharge have
higher decit volumes. This wetness effect was also concluded in
Section 4.1 and hypothesised in Section 1.
In Section 4.1 the ambiguous relation between N.GES and
H.MEAN was already noted, but with statistical analysis we can

0.0001672 ()

Signif. codes: 0 0.001 0.01 0.05 . 0.1 1.

5 did not yield a signicant model (not shown). In the stepwise

regression analysis adding the variables of model 5 one-by-one
resulted in a decreasing Akaike Information Criterion, so increasing
model performance. We can conclude that in our study area BFI is
made up of a combination of catchment characteristics related to
geology, land use and area. However, the information content of

(a)

10

20

500

15

15

Q_meandef

10

Brewa
Buwe
Dachst
Flysch
Gail
Gurk
Hoalp
Innv
Muhlv
Ozt
Waldv
Weivn

Q_meandef

Brewa
Buwe
Dachst
Flysch
Gail
Gurk
Hoalp
Innv
Muhlv
Ozt
Waldv
Weivn

1000

1500

20

(b)

500

2000

1000

1500

N.GES

2000

2500

3000

H.MEAN

2000

(c)

1500

1000

500

N.GES

Brewa
Buwe
Dachst
Flysch
Gail
Gurk
Hoalp
Innv
Muhlv
Ozt
Waldv
Weivn

500

1000

1500

2000

2500

3000

H.MEAN
Fig. 5. Relation between the average decit of droughts in discharge (Q_meandef) and (a) mean annual precipitation (N.GES), and (b) average elevation (H.MEAN), and (c) the
relation between N.GES and H.MEAN; colour coded per cluster.

A.F. Van Loon, G. Laaha / Journal of Hydrology 526 (2015) 314

11

Fig. 6. Example of the effect of a higher mean discharge on drought decit volume, (a) comparing two catchments in the Leitha cluster, and (b) comparing a catchment in the
Brewa and one in the Otz cluster.

Table 6
Linear models tested to explain the average decit volume of discharge droughts
(Q_meandef), with multiple regression and interaction terms, including their
statistical signicance.

N.GES
H.MEAN

Model 1

Model 2

Model 3

Model 4

5.54e 10
()

< 2e 16
()
3.64e 11
()

<2e 16
()
1.08e 11
()
0.0243 ()

1.904e 12
()

0.028516
()

Interaction (N.GES,
H.MEAN)

2.72e 08
()

Comparing models

Signif. codes: 0 0.001 0.01 0.05 . 0.1 1.

study if there is any additional information in H.MEAN. Linear

regressions models (Table 6) prove that both N.GES and H.MEAN
are important in modelling Q_meandef (models 13). The interaction between both is also signicant (model 4), but it does not
result in a better model (p = 0.029 instead of 1.9e 12). Since a
bimodal distribution is visible between N.GES and H.MEAN
(Fig. 5c), i.e. in some clusters precipitation does not increase with
altitude, we need a varying slope model to account for the confounding effects of altitude and precipitation. We grouped the
catchments in the southern Alps region that have a deviating
relation of precipitation with altitude (Ozt, Hoalp, Gail and Gurk;
Figs. 1 and 5c) and tted a different relation between Q_meandef
and N.GES for this selection and for the remaining catchments.
In Fig. 7 we see a clear effect of H.MEAN on Q_meandef (after
adjusting for precipitation effects). We also tried a model that
differentiates between the southern Alpine catchments and the
rest in terms of H.MEAN, but that did not give satisfactory results
(not shown).
From this analysis we can conclude that, additional to the precipitation effect mentioned above, the elevation of a catchment
plays a large role in determining the average drought decit. The
higher the catchment, the larger is the decit volume. This can

be explained by the stronger seasonality of ows in higher altitudes, i.e. alpine areas have stronger seasonal regimes compared
to lowland areas (Weingartner et al., 2013). Due to snow accumulation in winter and snow melt and glacier melt in summer,
streamow is concentrated in a short season. This results in seasonally high thresholds and hence high decit volumes. In the
example in Fig. 6b, the hydrographs of two catchments with similar decit volume (on average 16 and 14 mm; see Tables 2 and 3)
show very different drought behaviour. In Brewa severe droughts
can occur in any season and decit volumes are high because of
the overall high threshold (caused by high precipitation on the
north side of the Alps), whereas in Otz severe droughts only occur
in the summer half year because of zero ows in winter and decit
volumes are high because of the seasonally high threshold (caused
by concentration of streamow in a short season in the highest
part of the Alps, Weingartner et al., 2013). According to the hydrological drought typology, the most severe droughts in Brewa are
classical rainfall decit droughts (Van Loon and Van Lanen, 2012)
and the most severe droughts in Otz are rain-to-snow-season
droughts, snowmelt droughts and glaciermelt droughts (Van Loon
and Van Lanen, 2012; Van Loon et al., 2014). This example underlines the fact that precipitation and altitude are different effects in
generating high decit volumes. The higher the precipitation, the
larger the possible deviations from normal, and therefore the
higher the decit volumes. And the higher the elevation, the more
the ow is concentrated in a short season and the higher the decit
volumes.
From this statistical analysis we can conclude that for hydrological drought duration catchment storage (various catchment variables with a combined effect represented by BFI) is dominant
and climate plays a role through the duration of dry spells. For
hydrological drought decit we found that catchment wetness
(depending on mean annual precipitation) and seasonality of the
regime (depending on elevation) are of equal importance. We did
the same analysis on the maximum and standard deviation of
drought duration and decit, but that did not change our
conclusions.

12

A.F. Van Loon, G. Laaha / Journal of Hydrology 526 (2015) 314

H.MEAN effect plot

hotspot.c3*N.GES effect plot

500 1000 1500 2000

hotspot.c3 : 0

18

hotspot.c3 : 1

40
16

30

12

Q_meandef

Q_meandef

14

10

20

10

6
0
4

500

1000

1500

2000

2500

3000

H.MEAN

500 1000 1500 2000

N.GES

Fig. 7. Varying slope model of Q_meandef vs. H.MEAN and N.GES (subdivided between Southern Alps catchments and the rest); contribution of each regression term to the
prediction.

5. Discussion
5.1. Data and methods
In this study, the variable threshold level method was used for
drought analysis. As mentioned before (Section 1) standardised
indices like SPI were not applied because they cannot provide
information on drought decit volumes. Instead of using a variable
threshold we could also have chosen a xed threshold, but
the yearly recurring winter low ows in the Alpine catchments
(Section 2) should not be reported as drought because they are
not a deviation from normal conditions (Van Loon and Van
Lanen, 2012; Van Loon, 2013). Anomalies in the high ow season
can be important in water resources management (hydropower
production), so therefore we decided to use a variable threshold
level, as was done previously in many other studies (e.g. Stahl,
2001; Nyabeze, 2004; Hirabayashi et al., 2008; Vidal et al., 2010;
Hannaford et al., 2011; Prudhomme et al., 2011; Van Huijgevoort
et al., 2013; Parry et al., 2012).
Pooling with a 30-days moving average was done to merge
dependent drought events. This pooling allowed us to focus on
the longer-term water shortages. We estimate the effect of the
pooling method on our conclusions to be negligible, because we
performed the same correlation analysis on drought events without pooling and obtained the same results. Also for the non-pooled
drought events catchment storage was most important for drought
duration, whereas mean annual precipitation and elevation played
a major role for drought decit. The statistical relations were only
slightly weaker.
5.2. Drought duration and decit
The relation between hydrological drought duration and BFI
(baseow index) is conrmed in earlier studies (Clausen and
Pearson, 1995; Zaidman et al., 2002; Fendekov and Fendek,
2012; Van Huijgevoort, 2014). In our study we found BFI to be a
proxy for the combination of a number of catchment characteristics indicative of storage. This is in accordance with Peters et al.

(2003), Peters et al. (2005), Bloomeld et al. (2009) and Salinas

et al. (2013). Bloomeld et al. (2009) showed that geology is a
key determinant of BFI. Although other factors play a role as well
(Tetzlaff et al., 2008), it is not possible to say that one parameter
(e.g. a soil index or the catchment area) is responsible for all the
variation in BFI (Bloomeld et al., 2009). The statistical analysis
presented in this paper conrmed that a combination of a high
number of catchment characteristics is needed to obtain the same
effect on drought duration as BFI.
The governing factors of drought decit are investigated in
some studies (Clausen and Pearson, 1995; Kjeldsen et al., 2000;
Van Lanen et al., 2013) and in all of those the decit was standardised to remove the effect of differences in catchment wetness. So,
although the relationship between drought decit and mean
annual precipitation was never investigated explicitly, the standardisation is an indication that mean annual precipitation does
play an important role. We also tested the relation between standardised decit and climate and catchment characteristics and
found that the effect of climate was completely removed by the
standardisation and only catchment control (BFI) remained.
5.3. General applicability of the results
Since this research is based on a specic study area, its conclusions might not be valid for other regions. The Greater Alpine
Region was chosen because of, on the one hand, the wealth of
hydrometeorological data and catchment descriptors available
and, on the other hand, the large variability in climate and catchment characteristics.
Within our study region differences in temperature and precipitation can be found related to the presence of mountains (decreasing temperature with altitude, increasing precipitation with
altitude except in the rain shadow, and increasing aridity towards
the east), inuencing also the accumulation of snow. However,
these differences are relatively minor as almost the entire study
region has a continental climate type and is governed by the
same weather pattern with large scale depressions moving in from
the north-west creating comparable meteorological drought

A.F. Van Loon, G. Laaha / Journal of Hydrology 526 (2015) 314

conditions. If we would compare this region to a region with a

completely different climate, for example much drier and governed
by multi-year weather patterns like in Australia, then we would
expect differences in drought duration to be more determined by
these differences in climate. However, when comparing catchments within that drier climate region again catchment characteristics will be dominant in determining hydrological drought
duration. Hence, our results are assumed to be applicable to
regions that are relatively uniform in climate. This reconciles our
results with the hypothesis of Tallaksen and Hisdal (1997) mentioned in Section 1, because in that paper the authors are probably
referring to the effects of climate at a larger scale.
The effect of elevation is not purely catchment control,
because it is actually related to mean annual temperature, i.e.
the lower the temperature, the higher the glacier cover and snow
accumulation, the higher the seasonality, the higher the decit
volume. Because it is related to temperature, the effect of altitude
is different in different climates. For example, in warmer climates
(e.g. in the tropics) the relation with altitude is different since
snow and glaciers generally occur at higher altitudes than in temperate and continental climates. In regions without snow and glaciers this effect of altitude might be totally absent, because the
concentration of streamow in a short season does not occur
(Weingartner et al., 2013). The effect of temperature also means
that on a cross-section from south to north on the Northern
Hemisphere there might be a relation between drought decit
and latitude.
Consequently, the conclusion of what is the dominant factor in
determining hydrological drought duration and decit is highly
dependent on scale. On a global scale, drought duration might be
more related to climate than to catchment control and the effect
of altitude on drought decit might not be obvious. However, on
the scale that water resources management takes place, namely a
regional to national scale, the climate is assumed to be governed
by the same weather generating mechanism and therefore relatively uniform, while geology, soil, land use and other catchment
characteristics vary more on these smaller scales.
Climate change may lead to non-stationary relationships
between hydrological variables. In this paper, we analysed average
relationships over a standard period as is common for low ow
regionalisation studies. This allows comparison between stations
even in case of weak non-stationarity. An analysis of possible
future changes in the found relationships would be a logical next
step after this research. This would, however, require much longer
records than usually available.
6. Conclusions
Drought analysis and statistical analysis on an extensive
Austrian dataset, combining hydrometeorological data from a high
number of catchments with thematic data on climate and catchment characteristics, prove that hydrological drought duration
and decit do not have the same governing factors. Hydrological
drought duration is determined by BFI (combining information
on storage in the catchment related to a number of different catchment characteristics) and the duration of dry spells in precipitation
(that increases with increasing aridity). BFI is most important, so
catchment control on drought duration is dominant. Hydrological
drought decit is determined by mean annual precipitation and
elevation, which both are related to climate control. The additional
effect of elevation is attributed to the concentration of discharge in
a short season (summer) with rainfall, snow melt and glacier melt
in highly seasonal regimes.
This information is relevant for sectors that experience different
drought impacts. Based on their climate and catchment characteristics catchments can be selected that are more sensitive to severe

13

drought in terms of drought duration and catchments that are

more sensitive to severe drought in terms of drought decit.
Acknowledgements
This research was undertaken as part of the European Union
(FP7) funded Integrated Project DROUGHT-R& SPI (Contract No.
282769). It is part of the programme of the Wageningen Institute
for Environment and Climate Research (WIMEK-SENSE) and it supports the work of the UNESCO-IHP VIII FRIEND programme. The
authors would like to thank the Austrian Climate Research
Program (ACRP Project B060362 CILFAD) for nancial support.
Analyses are based on discharge records of the Hydrographical
Service of Austria (HZB), and precipitation and temperature
records of the Central Institute for Meteorology and Geodynamics
(ZAMG). We thank Klaus Haslinger (Central Institute for Meteorology and Geodynamics, Austria) for his support in selecting and
interpreting meteorological data of the study catchments and our
colleagues Paul Torfs (Wageningen University) and Daniel Kofer
(BOKU) for assistance with the data analysis.
References
Akaike, H., 1974. A new look at the statistical model identication. IEEE Trans.
Automatic Control 19 (6), 716723. http://dx.doi.org/10.1109/TAC.1974.
1100705.
Beyene, B.S., Van Loon, A.F., Van Lanen, H.A.J., Torfs, P.J.J.F., 2014. Investigation of
variable threshold level approaches for hydrological drought identication.
Hydrol. Earth Syst. Sci. Discuss. 11, 1276512797. http://dx.doi.org/10.5194/
hessd-11-12765-2014.
Bloomeld, J.P., Marchant, B.P., 2013. Analysis of groundwater drought building on
the standardised precipitation index approach. Hydrol. Earth Syst. Sci. 17 (12),
47694787. http://dx.doi.org/10.5194/hess-17-4769-2013.
Bloomeld, J.P., Allen, D.J., Grifths, K.J., 2009. Examining geological controls on
baseow index (b) using regression analysis: an illustration from the thames
basin, UK. J. Hydrol. 373 (12), 164176. http://dx.doi.org/10.1016/j.jhydrol.
2009.04.025.
Bond, N., Lake, P., Arthington, A., 2008. The impacts of drought on freshwater
ecosystems: an Australian perspective. Hydrobiologia 600 (1), 316. http://
dx.doi.org/10.1007/s10750-008-9326-z.
Carroll, N., Frijters, P., Shields, M.A., 2009. Quantifying the costs of drought: new
evidence from life satisfaction data. J. Population Econ. 22 (2), 445461. http://
dx.doi.org/10.1007/s00148-007-0174-3.
Choat, B., Jansen, S., Brodribb, T.J., Cochard, H., Delzon, S., Bhaskar, R., Bucci, S.J.,
Feild, T.S., Gleason, S.M., Hacke, U.G., Jacobsen, A.L., Lens, F., Maherali, H.,
Martinez-Vilalta, J., Mayr, S., Mencuccini, M., Mitchell, P.J., Nardini, A.,
Pittermann, J., Pratt, R.B., Sperry, J.S., Westoby, M., Wright, I.J., Zanne, A.E.,
2012. Global convergence in the vulnerability of forests to drought. Nature 491
(7426), 752755. http://dx.doi.org/10.1038/nature11688.
Clausen, B., Pearson, C., 1995. Regional frequency analysis of annual maximum
streamow drought. J. Hydrol. 173 (14), 111130, doi:http://dx.doi.org/
10.1016/0022-1694(95)02713-Y.
Dracup, J.A., Kil Seong, L., Paulson Jr, E.G., 1980. On the statistical characteristics of
drought events. Water Resour. Res. 16 (2), 289296.
Fendekov, M., Fendek, M., 2012. Groundwater drought in the Nitra River Basin
identication and classication. J. Hydrol. Hydromech. 60, 185.
Fleig, A.K., Tallaksen, L.M., Hisdal, H., Demuth, S., 2006. A global evaluation of
streamow drought characteristics. Hydrol. Earth Syst. Sci. 10 (4), 535552.
http://dx.doi.org/10.5194/hess-10-535-2006.
Fox, J., 2003. Effect displays in r for generalised linear models. J. Stat. Softw. 8 (15),
127.
Gal, L., Szolgay, J., Kohnov, S., Parajka, J., Merz, R., Viglione, A., Blschl, G., 2012.
Flood timescales: understanding the interplay of climate and catchment
processes through comparative hydrology. Water Resour. Res. 48 (4). http://
dx.doi.org/10.1029/2011WR011509.
Garcia-Herrera, R., Daz, J., Trigo, R.M., Luterbacher, J., Fischer, E.M., 2010. A review of
the european summer heat wave of 2003. Crit. Rev. Environ. Sci. Technol. 40 (4),
267306.
Hannaford, J., Lloyd-Hughes, B., Keef, C., Parry, S., Prudhomme, C., 2011. Examining
the large-scale spatial coherence of European drought using regional indicators
of precipitation and streamow decit. Hydrol. Process. 25 (7), 11461162.
http://dx.doi.org/10.1002/hyp.7725.
Haslinger, K., Kofer, D., Schner, W., Laaha, G., 2014. Exploring the link between
meteorological drought and streamow: effects of climate-catchment
interaction. Water Resour. Res. 50 (3), 24682487. http://dx.doi.org/10.1002/
2013WR015051.
Hayes, M., Svoboda, M., Wall, N., Widhalm, M., 2010. The lincoln declaration on
drought indices: universal meteorological drought index recommended. Bull.
Am. Meteor. Soc. 92 (4), 485488. http://dx.doi.org/10.1175/2010BAMS3103.1.

14

A.F. Van Loon, G. Laaha / Journal of Hydrology 526 (2015) 314

Hirabayashi, Y., Kanaae, S., Emori, S., Oki, T., Kimoto, M., 2008. Global projections of
changing risks of oods and droughts in a changing climate. Hydrol. Sci. J. 53
(4), 754772. http://dx.doi.org/10.1623/hysj.53.4.754.
Hisdal, H., Tallaksen, L.M., Clausen, B., Peters, E., Gustard, A., 2004. Hydrological
Drought Characteristics. Elsevier Science B.V., Amsterdam, the Netherlands,
Developments in Water Science, 48, 2004, Chapter 5, pp. 139198.
Hsiang, S.M., Burke, M., Miguel, E. Quantifying the inuence of climate on human
conict, Science 341 (6151). arXiv:http://www.sciencemag.org/content/341/
6151/1235367.full.pdf, http://dx.doi.org/10.1126/science.1235367.
Joetzjer, E., Douville, H., Delire, C., Ciais, P., Decharme, B., Tyteca, S., 2013.
Hydrologic benchmarking of meteorological drought indices at interannual to
climate change timescales: a case study over the amazon and mississippi river
basins. Hydrol. Earth Syst. Sci. 17 (12), 48854895. http://dx.doi.org/10.5194/
hess-17-4885-2013.
Jonsdottir, H., Eliasson, J., Madsen, H., 2005. Assessment of serious water shortage in
the icelandic water resource system. Phys. Chemis. Earth, Parts A/B/C 30 (67),
420425. http://dx.doi.org/10.1016/j.pce.2005.06.007, integrated Water
Resource Assessment.
Kim, T., Valds, J., Yoo, C., 2003. Nonparametric approach for estimating return
periods of droughts in arid regions. J. Hydrol. Eng. 8 (5), 237246. http://
dx.doi.org/10.1061/(ASCE)1084-0699(2003)8:5(237).
Kjeldsen, T.R., Lundorf, A., Rosbjerg, D., 2000. Use of a two-component exponential
distribution in partial duration modelling of hydrological droughts in
zimbabwean rivers. Hydrol. Sci. J. J. Sci. Hydrol. 45 (2), 285298.
Kofer, D., Laaha, G. Lfstat-low-ow analysis in r. EGU General Assembly
Conference Abstracts 15, 2013
Laaha, G., Blschl, G., 2006. Seasonality indices for regionalizing low ows. Hydrol.
Process. 20 (18), 38513878.
Lake, P.S., 2011. Drought and Aquatic Ecosystems: Effects and Responses. John
Wiley & Sons.
Lewis, S.L., Brando, P.M., Phillips, O.L., van der Heijden, G.M.F., Nepstad, D., 2011.
The 2010 Amazon drought. Science 331 (6017), 554. http://dx.doi.org/10.1126/
science.1200807.
McKee, T., Doesken, N., Kleist, J., 1993. The relationship of drought frequency and
duration to time scales. In: Eight Conference on Applied Climatology, 1722
January 1993, Anaheim, California, Amer. Meteor. Soc., pp. 179184. <http://
ccc.atmos.colostate.edu/relationshipofdroughtfrequency.pdf> (last accessed
16.07.12).
Mishra, A.K., Singh, V.P., 2010. A review of drought concepts. J. Hydrol. 391 (1-2),
202216. http://dx.doi.org/10.1016/j.jhydrol.2010.07.012.
Mishra, A., Singh, V., Desai, V., 2009. Drought characterization: a probabilistic
approach. Stochastic Environ. Res. Risk Assessm. 23 (1), 4155.
Nyabeze, W.R., 2004. Estimating and interpreting hydrological drought indices
using a selected catchment in Zimbabwe. Phys. Chem. Earth 29 (1518), 1173
1180.
Parry, S., Hannaford, J., Lloyd-Hughes, B., Prudhomme, C., 2012. Multi-year droughts
in Europe: analysis of development and causes. Hydrol. Res. 43 (5), 689706.
http://dx.doi.org/10.2166/nh.2012.024.
Peters, E., Torfs, P.J.J.F., van Lanen, H.A.J., Bier, G., 2003. Propagation of drought
through groundwater a new approach using linear reservoir theory. Hydrol.
Process. 17 (15), 30233040. http://dx.doi.org/10.1002/hyp.1274.
Peters, E., van Lanen, H.A.J., Torfs, P.J.J.F., Bier, G., 2005. Drought in groundwater
drought distribution and performance indicators. J. Hydrol. 306 (14), 302317.
Peters, E., Bier, G., van Lanen, H.A.J., Torfs, P.J.J.F., 2006. Propagation and spatial
distribution of drought in a groundwater catchment. J. Hydrol. 321 (1/4), 257
275. http://dx.doi.org/10.1016/j.jhydrol.2005.08.004.
Prudhomme, C., Parry, S., Hannaford, J., Clark, D.B., Hagemann, S., Voss, F., 2011.
How well do large-scale models reproduce regional hydrological extremes in
Europe? J. Hydrometeorol. 12 (0), 11811204. http://dx.doi.org/10.1175/
2011JHM1387.1.
Rossi, G., Caporali, E., Garrote, L., 2012. Denition of risk indicators for reservoirs
management optimization. Water Resour. Manage. 26 (4), 981996. http://
dx.doi.org/10.1007/s11269-011-9842-x.
Salinas, J.L., Laaha, G., Rogger, M., Parajka, J., Viglione, A., Sivapalan, M., Blschl, G.,
2013. Comparative assessment of predictions in ungauged basins part 2: ood
and low ow studies. Hydrol. Earth Syst. Sci. 17 (7), 26372652. http://
dx.doi.org/10.5194/hess-17-2637-2013.
Shefeld, J., Wood, E., 2011. Drought; Past Problems and Future Scenarios.
Earthscan, London, Washington DC.
Shiau, J., Shen, H., 2001. Recurrence analysis of hydrologic droughts of differing
severity. J. Water Resour. Plann. Manage. 127 (1), 3040. http://dx.doi.org/
10.1061/(ASCE)0733-9496(2001)127:1(30).
Stagge, J., Tallaksen, L., Gudmundsson, L., Van Loon, A.F., Stahl, K., submitted for
publication. Candidate distributions for climatological drought indices (SPI and
SPEI). Int. J. Climatol.
Stahl, K., 2001. Hydrological Drought a Study across Europe. Ph.D. thesis, AlbertLudwigs-Universitt, Freiburg, Germany.
Tallaksen, L.M., Hisdal, H., 1997. Regional analysis of extreme streamow drought
duration and decit volume, FRIEND 97 regional hydrology: concepts and

models for sustainable water resource management. In: Proc. International

Conference, Postojna, vol. 246, pp. 141150.
Tallaksen, L.M., Van Lanen, H.A.J. (Eds.), 2004. Hydrological drought: processes and
estimation methods for streamow and groundwater. In: Developments in
Water Science; 48, The Netherlands, Elsevier Science B.V., Amsterdam, The
Netherlands.
Tetzlaff, D., McDonnell, J.J., Uhlenbrook, S., McGuire, K.J., Bogaart, P.W., Naef, F.,
Baird, A.J., Dunn, S.M., Soulsby, C., 2008. Conceptualizing catchment processes:
simply too complex? Hydrol. Process. 22 (11), 17271730. http://dx.doi.org/
10.1002/hyp.7069.
Tsakiris, G., Nalbantis, I., Vangelis, H., Verbeiren, B., Huysmans, M., Tychon, B.,
Jacquemin, I., Canters, F., Vanderhaegen, S., Engelen, G., Poelmans, L., Becker, P.,
Batelaan, O., 2013. A system-based paradigm of drought analysis for operational
management. Water Resour. Manage. 27 (15), 52815297. http://dx.doi.org/
10.1007/s11269-013-0471-4.
Van Huijgevoort, M., 2014. Hydrological drought: characterisation and
representation in large-scale models. Ph.D. thesis, Wageningen University, the
Netherlands.
Van Huijgevoort, M.H.J., Hazenberg, P., van Lanen, H.A.J., Teuling, A.J., Clark, D.B.,
Folwell, S., Gosling, S.N., Hanasaki, N., Heinke, J., Koirala, S., Stacke, T., Voss, F.,
Shefeld, J., Uijlenhoet, R., 2013. Global multimodel analysis of drought in
runoff for the second half of the twentieth century. J. Hydrometeor 14, 1535
1552. http://dx.doi.org/10.1175/JHM-D-12-0186.1.
Van Lanen, H.A.J., 2006. Drought propagation through the hydrological cycle. In:
Demuth, S., Gustard, A., Planos, E., Scatena, F., Servat, E. (Eds.), Climate
Variability and Change Hydrological Inpacts, vol. 308. IAHS-AISH Publication,
IAHS Press, Wallingford, UK, pp. 122127.
Van Lanen, H.A.J., Wanders, N., Tallaksen, L.M., Van Loon, A.F., 2013. Hydrological
drought across the world: impact of climate and physical catchment structure.
Hydrol. Earth Syst. Sci. 17, 17151732. http://dx.doi.org/10.5194/hess-171715-2013.
Van Loon, A.F., 2013. On the propagation of drought. How climate and catchment
characteristics inuence hydrological drought development and recovery. Ph.D.
thesis, Wageningen University. <http://edepot.wur.nl/249786> (last accessed
28.05.13).
Van Loon, A.F., Van Lanen, H.A.J., 2012. A process-based typology of hydrological
drought. Hydrol. Earth Syst. Sci. 16 (7), 19151946. http://dx.doi.org/10.5194/
hess-16-1915-2012.
Van Loon, A.F., Tijdeman, E., Wanders, N., Van Lanen, H., Teuling, A., Uijlenhoet, R.,
2014. How climate seasonality modies drought duration and decit. J.
Geophys. Res.: Atmos. 119 (8), 46404656. http://dx.doi.org/10.1002/
2013JD020383.
Van Loon, A.F., Ploum, S.W., Parajka, J., Fleig, A.K., Garnier, E., Laaha, G., Van Lanen,
H.A.J., 2014. Hydrological drought typology: cold climate drought types and
associated societal impacts. Hydr. Earth Syst. Sci. Discuss. 11, 1046510514.
http://dx.doi.org/10.5194/hessd-11-10465-2014.
Van Vliet, M.T.H., Yearsley, J.R., Ludwig, F., Vogele, S., Lettenmaier, D.P., Kabat, P.,
2012. Vulnerability of US and European electricity supply to climate change.
Nature Clim. Change 2 (9), 676681. http://dx.doi.org/10.1038/nclimate1546.
Vicente-Serrano, S.M., Beguera, S., Lpez-Moreno, J.I., 2009. A multiscalar drought
index sensitive to global warming: the standardized precipitation
evapotranspiration index. J. Climate 23 (7), 16961718. http://dx.doi.org/
10.1175/2009JCLI2909.1.
Vidal, J.-P., Martin, E., Franchistguy, L., Habets, F., Soubeyroux, J.-M., Blanchard, M.,
Baillon, M., 2010. Multilevel and multiscale drought reanalysis over France with
the Safran-Isba-Modcou hydrometeorological suite. Hydrol. Earth Syst. Sci. 14
(3), 459478. http://dx.doi.org/10.5194/hess-14-459-2010.
Weingartner, R., Blschl, G., Hannah, D., Marks, D., Parajka, J., Pearson, C., Rogger, M.,
Salinas, J.L., Sauquet, E., Srikanthan, R., Thompson, S., Viglione, A., 2013.
Prediction of Seasonal Runoff in Ungauged Basins, Cambridge, pp. 102134
(Chapter 6).
Wilhite, D. (Ed.), 2000. Drought: A Global Assessment, vols. I &II. Routledge Hazards
and Disasters Series, Routledge, London.
Wilhite, D.A., Glantz, M.H., 1985. Understanding the drought phenomenon: the role
of denitions. Water Int. 10 (3), 111120.
WMO, 2008. Manual on low ow estimation and prediction. Operational Hydrology
Report No.50 WMO-No. 1029. p. 136.
Wong, G., Van Lanen, H.A.J., Torfs, P.J.J.F., 2013. Probabilistic analysis of hydrological
drought characteristics using meteorological drought. Hydrol. Sci. J. 58 (2), 253
270.
Woo, M.-K., Tariiule, A., 1994. Streamow droughts of northern Nigerian rivers.
Hydrol. Sci. J. 39 (1), 1934.
Zaidman, M.D., Rees, H.G., Young, A.R., 2002. Spatio-temporal development of
streamow droughts in northwest Europe. Hydrol. Earth Syst. Sci. 6 (4), 733
751.
Zelenhasic, E., Salvai, A., 1987. A method of streamow drought analysis. Water
Resour. Res. 23 (1), 156168. http://dx.doi.org/10.1029/WR023i001p00156.